Elon Musk’s AI chatbot Grok restricted access to a much-criticized deepfake generator on social media platform X following a surge of users creating nonconsensual nude images.

Grok now says that “image generation is currently limited to paying subscribers” as “this helps ensure responsible use while we continue refining things,” citing “recent issues and improvements to safeguards.”

The chatbot has drawn scrutiny from regulators and politicians across Europe after it enabled users to manipulate pictures posted online into a series of deepfakes, including depictions of undressed minors and public figures.

Swedish Deputy Prime Minister Ebba Busch was one of the latest high-profile victims. That post is no longer visible. “This post from grok has been withheld in [the] European Union based on local law(s),” it is now labeled.

The European Commission ordered X on Thursday to retain “all internal documents and data relating to Grok” — an escalation of the ongoing investigation into X’s content moderation policies — after calling the nonconsensual, sexually explicit deepfakes “illegal,” “appalling” and “disgusting.”

X did not respond to a request for comment in time for publication.